Differences in Delineation Guidelines for Head & Neck Cancer Result in Inconsistentreported

Differences in delineation guidelines for head & neck cancer result in inconsistentreported dose and corresponding NTCP-values

Brouwer et al. 2013

Table S1. Overview of delineation guidelines for swallowing organs at risk (SWOARs) and the clustering into delineation groups (DGs) with corresponding definitions. The number of different definitions and therefore the number of DGs varies per structure. SWOARs used as input parameters of NTCP models [12] are marked by *.

SWOAR / Delineation groups (DGs) of corresponding definitions
DG1 / DG2 / DG3 / DG4 / DG5
Superior PCM* / Christianen
Cranial: caudal tips of pterygoid plates
Caudal: lower edge of C2 / Dirix, Caglar, Caudell, Feng
Cranial: caudal tips of pterygoid plates
Caudal: superior end of hyoid bone / Levendag
Cranial: mid C2
Caudal: upper C3 / Bhide
Cranial: base of the skull
Caudal: upper edge of hyoid bone
Middle PCM* / Christianen
Cranial: upper edge of C3
Caudal: lower edge of hyoid bone / Bhide, Dirix, Caglar, Caudell, Feng,
Cranial: upper edge of hyoid bone
Caudal: lower edge of hyoid bone / Levendag
Cranial: upper C3
Caudal: lower edge of hyoid bone
Inferior PCM / Christianen
Cranial: first slice caudal to the lower edge of hyoid bone
Caudal: lower edge of arythenoid cartilages / Bhide, Dirix, Caglar, Caudell
Cranial: lower edge of hyoid bone
Caudal: lower edge of cricoid cartilage
Total PCM / Christianen
Cranial: caudal tips of pterygoid plates
Caudal: lower edge of arythenoid cartilages / Dirix, Cagar, Caudell, Feng, Li
Cranial: caudal tips of pterygoid plates
Caudal: lower edge of cricoid cartilage / Levendag
Cranial: mid C2
Caudal: lower edge of cricoids cartilage / Bhide
Cranial: base of the skull
Caudal:lower edge of cricoid cartilage / Jensen
Cranial: lower part of transverse process of C2
Caudal: top of cricoid cartilage
Supgraglottic larynx* / Christianen
Cranial: tip of epiglottis
Caudal: first slice cranial to upper edge of arytenoid cartilages / Dirix, Jensen
Cranial: top of piriform sinus and aryepiglottic fold
Caudal: upper edge of cricoid cartilage
Glottic larynx / Christianen
Cranial: upper edge of arythenoid cartilages
Caudal: lower edge of cricoid cartilage / Dirix, Jensen
Cranial: upper edge of cricoid cartilage
Caudal: lower edge of cricoid cartilage
Larynx / Christianen
Cranial: tip of epiglottis
Caudal: lower edge of cricoid cartilage / Dirix, Jensen
Cranial: top of piriform sinus and aryepiglottic fold
Caudal: lower edge of cricoid cartilage / Caglar
Cranial: upper edge of thyroid cartilage
Caudal: upper edge of cricoid / Caudell
Cranial: epiglottis
Caudal: vocal cords
Cricopharyngeus / Christianen
Cranial: first slice caudal to arytenoid cartilages
Caudal: lower edge of cricoid cartilages / Li
Cranial: caudal cricoid
Caudal: first tracheal ring, 1 cm caudal from cranial border / Levendag
Cranial: caudal edge cricoid
Caudal: lower border First trachea ring
Cricopharyngeus & EIM / Christianen
Cranial: first slice caudal to arytenoid cartilages
Caudal: 1 cm caudal to the superior border / Dirix
Cranial: lower edge cricoid cartilage
Caudal: upper edge of trachea / Levendag
Cranial: caudal edge cricoid
Caudal: 1 cm caudal from lower border first trachea ring

PCM= pharyngeal constrictor muscles, EIM= esophageal inlet muscle

Table S2.Volume of swallowing organs at risk (SWOARs) for the different delineation groups (DGs). PCM = pharyngeal constrictor muscles. EIM = esophageal inlet muscle.

SWOAR / Volume (cc)*
DG1 / DG2 / DG3 / DG4 / DG5 / -value**
Total PCM / 20.8 (4.9) / 29.9 (6.9) / 22.4 (6.2) / 26.7 (7.2) / 9.8 (3.6) / 0.000
Superior PCM / 7.6 (1.8) / 12.5 (3.6) / 4.5 (1.4) / 9.9 (3.8) / - / 0.000
Middle PCM / 4.3 (2.2) / 2.6 (0.9) / 5.3 (2.8) / - / - / 0.000
Inferior PCM / 4.8 (1.3) / 10.1 (2.8) / - / - / - / 0.000
Larynx / 21.1 (6.2) / 9.0 (4.0) / 18.7 (5.5) / 16.2 (5.9) / - / 0.000
Supraglottic Larynx / 10.9 (3.5) / 9.2 (4.1) / - / - / - / 0.041
Glottic Larynx / 7.4 (2.9) / 5.1 (1.8) / - / - / - / 0.000
Cricopharyngeus / 3.7 (1.4) / 1.8 (0.6) / 0.8 (0.3) / - / - / 0.000
Cricopharyngeus & EIM / 5.9 (1.9) / 0.6 (0.4) / 3.0 (0.9) / - / - / 0.000
* Variables are denoted as mean (standard deviation). ** Group differences were tested with Friedman’s ANOVA.

Figure S1. Boxplot of the Concordance Index (CI) of the DGs in relation to DG1 in all 29 patients. No data for the cricopharyngeal muscle as a separate structure is presented since the CI with DG1 was 0 for all patients. sup= superior, med= middle, inf= inferior and total= total structure.
Boxplot parameters: The bottom and top of the box are the 25th and 75th percentile (the lower and upper quartiles, respectively) of the data. The error bars represent the minimum and maximum values of the data.

Figure S2. Mean dose to the swallowing organs at risk according to the different delineation groups (DGs), sorted on largest difference (left-right). PCM= pharyngeal constrictor muscle, CP= cricopharyngeal muscle, EIM = esophageal inlet muscle.

NTCP Calculation

NTCP calculations were based on a standard model of El Naqa.[1] Its mathematical definition was described in (1). The s-value differed for each model and its value was a summation of the dose and weights for the OARs for that specific model. The resulting NTCP gave the chance on complications for the OARS as a value between 0 and 1.

(1)

In which

The parameters of the different models () are given in Table 2. The age-factor was 0 for patients of age 18-65 and 1 for patients of an age above 65. In terms of radiation technique a 0 was for 3D-CRT and a 1 for IMRT. The tumour site value was 1 when it was primary located at the oropharynx or the nasopharynx, otherwise it was 0.

Table S3.Overview of s-values for physician- and patient-rated swallowing dysfunction according to institutional prediction models.[2]

Swallowing dysfunction / -value
RTOG grade 2-4 / ) +

Liquid food /

Solid food /
Soft food /

Choking /

1. El Naqa I, Bradley J, Blanco AI, et al.Multivariable modeling of radiotherapy outcomes, including dose-volume and clinical factors. International journal of Radiation Oncology, Biology, Physics. 2006;64:1275–86.

2. Christianen MEMC, Schilstra C, Beetz I, et al.Predictive modelling for swallowing dysfunction after primary (chemo)radiation: results of a prospective observational study. Radiotherapy and Oncology. 2012;105:107–14.

Standard versus Swallowing sparing IMRT plans

This section shows that in standard as well as in swallowing sparing IMRT (when plans are optimized by reducing the dose to the SWOARs as well), deviations in NTCP-valuesremain if delineations of DG2 are used instead of DG1 (NTCP-model based on DG1).

In this example, we calculated the NTCP-value for grade II-IV dysphagia based on the mean dose to the superior PCM and supraglottic area for two different delineations (DG1 and DG2) (Figure S3).In addition, the treatment plan was optimized by reducing the dose to the SWOARs, resulting in plan DG1-optimized and plan DG2-optimized. The NTCP-values for the different planswere then calculated again (Table S4).

Figure S4 shows dose volume histogramsof the PCM superior for the standard and the DG2-optimized IMRT plan. For both the standard and optimized plan (solid and dashed lines, respectively), the dose to the PCM superior based on the DG2 delineation(in red) deviated from the dose according to the DG1 delineation (in green).

For the incorrect situation, the NTCP is calculated based on the dose according to the DG2 delineation (red lines in Figure S4).The corresponding NTCP-value for the DG2-optimized IMRT plan was9.3%, while the correct NTCP-value for this dose distribution, based on the DG1 delineation (dashed green dose-volume line in Figure S4),should be 6.1% (Table S4). This absolute difference of 3.3% in NTCP translates into a relative NTCP overestimation of 54%. Although in this case the absolute difference is relatively small, this example proves that optimized IMRT based on SWOARs do not automatically reduces the problem of under- or overestimation of NTCP-values.

So for both standard and swallowing sparing IMRT plans, similar errors are made if incorrect NTCP model parameters are used.

Figure S3. Delineations of the PCM superior ad supraglottic larynx (green= DG1, red= DG2) and PTV70 (purple).

Figure S4. Dose volume histograms (DVHs) of the PCM superior for two different IMRT plans (solid: standard IMRT, dashed: DG2-optimized IMRT) and the two different delineation guidelines (green: DG1, red: DG2). DG= delineation group.

TableS4. Dose and NTCP results for grade II-IV dysphagia, for three different treatment plans.

Standard IMRT / Swallowing Sparing IMRT
DG1-Optimized / DG2-Optimized
PCM superior MD for delineations of DG1 (Gy) / 33,2 / 20,9 / 19,6
PCM superior MD for delineations of DG2 (Gy) / 36,0 / 25,6 / 23,4
Supraglottic MD for delineations of DG1 (Gy) / 62,1 / 60,8 / 60,4
Supraglottic MD for delineations of DG2 (Gy) / 68,7 / 67,8 / 67,1
NTCP for delineations of DG1 (%) / 13,0 / 6,6 / 6,1
NTCP for delineations of DG2 (%) / 18,3 / - / 9,3
abs ΔNTCP(%) / 5,3 / - / -3,3
rel ΔNTCP(%) / 40% / - / -54%

PCM= pharyngeal constrictor muscles, MD= mean dose, DG= delineation group.
Green and red lines correspond to the dose volume histograms of Figure S4.